towards the dynamic semantic web

Nokia Internal Use Only 1

Towards the Dynamic Semantic Web

Ian Oliver

Nokia Research Centre

Helsinki, Finland

ruSmart 2008

St.Petersburg, Russia

5 September 2008


Contents

Part one

• History

·Web to Semantic Web

•Technologies

·Applications, Pages and Agents

Part two

•Semantic Computation

· Spaces and Agents

· Scalability

· What are Applications?

•Theoretical Underpinnings

· Graphs, Graph Structures

· Agents, Spaces

· Ontologies, Semantics

· Total Abstraction

•Demonstration


Part One


History

time 2010 1990/2000

usenet

gopher

ftp

uucp


History

time 2010 1990/2000

Web 1.0

usenet

gopher

ftp

uucp

www


History

time 2010 1990/2000

Web 1.0

usenet

gopher

ftp

uucp

Web 2.0

tagging

flickr

facebook

maps

blogging

www


History

time 2010 1990/2000

Web 1.0

usenet

gopher

ftp

uucp

Web 2.0

tagging

flickr

facebook

maps

Semantic Web

blogging

www


History

time 2010 1990/2000

Web 1.0

usenet

gopher

ftp

uucp

Web 2.0

tagging

flickr

facebook

maps

Semantic Web

Web 3.0

location

blogging

www


History

time 2010 1990/2000

Web 1.0

usenet

gopher

ftp

uucp

Web 2.0

tagging

flickr

facebook

maps

Semantic Web

blogging

Web 3.0

location

www


History

time 2010 1990/2000

Web 1.0

usenet

gopher

ftp

uucp

Web 2.0

tagging

flickr

facebook

maps

Semantic Web

blogging

Semantic Computation

Web 3.0

location

www


Technologies - Applications m

atu

rity

time 2010 1990/2000



atu

rity

time 2010 1990/2000

Device oriented

Symbian, Windows, Linux

Applications



atu

rity

time 2010 1990/2000

local/internet connectivity

PIM

Office

Device oriented


Applications



atu

rity

time 2010 1990/2000

Device oriented


Applications

PIM

Office


Page

Browser oriented

Firefox, IE, SOA, IM ...



atu

rity

time 2010 1990/2000

Device oriented


Applications

PIM

Office


Page

Lifeblog, Lifestyle,

Maps, Widsets, Applets

Browser oriented


location, internet enhanced,

content v presentation



atu

rity

time 2010 1990/2000

Device oriented


Applications

PIM

Office


Page



Browser oriented


Agent?



Space oriented

Sedvice, SmartSpaces ...



atu

rity

time 2010 1990/2000

Device oriented


Applications

PIM

Office


Page



Browser oriented


?

semantics, total integration, reasoning



Space oriented


Agent?



atu

rity

time 2010 1990/2000

Device oriented


Applications

PIM

Office


Page



Browser oriented


?

semantics, total integration, reasoning



Space oriented


Agent

Web 1.0

Web 2.0

Semantic Web Web 3.0

Semantic Computatio

n


Part Two


Web 1.x, 2.x, 3.x & Semantic Web Characteristics

Webs 1.0, 2.0, 3.0

• content oriented

·news, media

• user publishing

·user generated content

·personal content (gmail, flickr,

geotagging etc)

·folksonomies, tagging

• search

·Google, Yahoo

Semantic Web

• information oriented

• classification

• rise of the ontology

· strict and structured

· enables reasoning, “AI” etc

• global information

• internet of things

• The Giant Global Graph


Semantic Computation

Webs 1.0, 2.0

• content oriented

·news, media

• user publishing

·user generated content

·personal content (gmail, flickr,

geotagging etc)

·folksonomies, tagging

• search

·Google, Yahoo

Semantic Web

• information oriented

• classification

• rise of the ontology

· strict and structured

· enables reasoning, “AI” etc

• global information

• internet of things

• The Giant Global Graph

+ a Model of Computation


Semantic Computation – a definition

at least, an attempt at a definition...

“Semantic Computation takes the current Web x.y and Semantic Web

concepts and unifies them into a global, ubiquitous computation

framework that enables total integration of information in localised,

personal contexts...”


Semantic Computation - concepts

Return of the agent

• done before?

Return of the space

• done before?



Return of the agent

• done before?

Return of the space

• done before?

Yes, but...

• it failed...where are the ubiquitous

agents today?



Return of the agent

• done before?

Return of the space

• done before?

Yes, but...

• it failed...where are the ubiquitous

agents today?

• lack of infrastructure

• lack of ubiquitous computation

resources

• lack of representation formats

• lack of classification hierarchies

• lack of standardisation

• lack of global understanding of

semantics


Sedvice-M3

“An environment supporting an space and actor/agent-based model of

computation upon a Semantic Web substrate providing for

integration and interoperability between applications and devices”


Sedvice-M3 Philosophy

•space-based computing environment

•multiple, individual autonomous spaces

·local information, reasoning, logics, ontologies etc

·distributed information

·distributed deductive closure

·D(S1(Q)S2(Q))D(S1 (Q)) D(S2 (Q))

•information sharing

·RDF, Semantic Web

·ontologies, tagging, folksonomies

•applications

·constructed from agents

·autonomous, anonymous, distributed, mobile

·control-flow through ontological means

·no control flow !!??

·may be made ”outside” of the system via NoTA, UPnP, Webservices etc

·semi-structured information

·no strict ontology conformance

·inconsistent information allowed!

·free logics

·non-monotonic

·semantics of information, belief and truth maintenance responsibility of the reader (agent/actor)

·everything is information

·everything is first-order

·first-order policy, security, belief and trust structures


Sedvice-M3 Architecture

Connectivity

Information Store

adaptation

Storage

TCP/IP

listener

NoTA

listener

Bluetooth

listener

Assert

handler

Retract

handler

Query

handler

Subscribe

handler

Assert

engine

Subscribe

engine

Query

engine

Retract

engine

Scheduler

Dynamic

reasoners

Operation

rewriting

engine

Object

rewriting

engine

Transaction

processing

...

SSAP

handler

SOAP

handler

SyncML

handler...

Protocol

handlers

RDF storeLocal storage

External

storage

Object storage

Smart Space Node

communication media

Ontological Convenience Libraries/Functionality



Connectivity

Information Store

adaptation

Storage

TCP/IP

listener

NoTA

listener

Bluetooth

listener

Assert

handler

Retract

handler

Query

handler

Subscribe

handler

Assert

engine

Subscribe

engine

Query

engine

Retract

engine

Scheduler

Dynamic

reasoners

Operation

rewriting

engine

Object

rewriting

engine

Transaction

processing

...

SSAP

handler

SOAP

handler

SyncML

handler...

Protocol

handlers


External

storage

Object storage

Smart Space Node

communication media


Message

synthesizer

Router

Planner

(lookup)

Caching

Distribution



Connectivity

Information Store

adaptation

Storage

TCP/IP

listener

NoTA

listener

Bluetooth

listener

Assert

handler

Retract

handler

Query

handler

Subscribe

handler

Assert

engine

Subscribe

engine

Query

engine

Retract

engine

Scheduler

Dynamic

reasoners

Operation

rewriting

engine

Object

rewriting

engine

Transaction

processing

...

SSAP

handler

SOAP

handler

SyncML

handler...

Protocol

handlers


External

storage

Object storage

Smart Space Node

communication media


Message

synthesizer

Router

Planner

(lookup)

Caching

Distribution

Policy,Security and

Trust


Scalability

Dynamic Information

Structure

Free-form Databases

Traditional Fixed Schema Databases

Local Information

Sharing

Information Dynamicity

Sca

lab

ilit

y (

tod

ay

)

Mb

Gb

Tb

Eb

Pb


Scalability

Dynamic Information

Structure

Free-form Databases

Traditional Fixed Schema Databases

Local Information

Sharing

Information Dynamicity

Sca

lab

ilit

y (

tod

ay

)

Mb

Gb

Tb

Eb

Pb

Sedvice

Tripcom

Oracle

Google

Everybit

Yahoo

IBM DB/2


Application Construction

Highly Structured

Highly Unstructured




Highly Structured

Highly Unstructured


Yahoo Pipes

Traditional/Legacy Application Devleopment

Nokia Widsets

Traditional Agents

Sedvice-M3 Agents


Applications

Traditional applications

•monolithic

•single purpose

•difficult to expand and enhance

•fixed focus

•integration with other applications

·impossible in an ad hoc manner


Applications

Semantic Computation based applications

•appearance decided by user

•UI functionality

•individual agent-based parts

•information gathered from numerous

sources

•reasoning about information

·eg: weather reports as email

·friends = contacts

·locations = contacts

·etc

•functionality is emergent


Applications


•granularity can be extremely fine


Applications


•each individual agent can exist outside of

the hosting UI concept

·emergent functionality

•each agent manages its “non-exclusive”

area of expertise and information


Theoretical Underpinnings - Graph

Information is a (directed) graph

·RDF + Reasoning

·RDF++ / Wilbur

·Everything is first-class


Theoretical Underpinnings - Graph

Within a graph information may be

typed according to some ontology

or, tagged according to some

folksonomy

or, both

type

type type

tag tag

tag

tag


Theoretical Underpinnings – Graph Structures

RDF Graph = mathematical graph +

additional constraints and deduction

· , xX,XY,YZ XZ x Z

· these rules can be modified by

the space’s logic

• Deduction is performed at query-

time, ie: dyamically

· some spaces might perform

deduction at insert-time



Two extremes:

RDF Triple

RDF Graph



Finer grained mechanisms necessary

RDF Triple

RDF Graph

Molecule Subgraph Scope

}


Theoretical Underpinnings – Scopes & Reflection

First-order characteristics of scope

•Reflection

•Scopes can be composed

•Scopes require additional operators

(other than graph traversal)

· union

· intersection

· guards/pre-conditions

· etc

·RDF as its own programming

language as well as representation?


Theoretical Underpinnings – Intentional Semantics

Interpretation to some semantic

grounding is made on a per agent

basis

ontologies and folksonomies provide

assistance only...

type

type type

tag tag

tag

tag

agent1 agent2

semantic domains

interprets interprets

grounding grounding


Theoretical Underpinnings - Intentional Semantics

Interpretation to some semantic

grounding is made on a per agent

basis

ontologies and folksonomies provide

assistance only...

...which helps in ensuring that a

common interpretation is made

· at least a common enough

interpretation

type

type type

tag tag

tag

tag

agent1 agent2

semantic domains


grounding grounding


Theoretical Underpinnings - Intentional Semantics

Semantics is intentional rather than

fixed.

The agent writing a given piece of

information provides meta-

information (type, tag, other

properties, relationships etc) to

indicate its intention how that piece

of information should be interpreted

cf: duck-typing, mixins, multiple-

inheritance, undecidablity, description

logic decision procedures etc...

type

type type

tag tag

tag

tag

agent1 agent2

semantic domains


grounding grounding


Theoretical Underpinnings – Agent Interoperability

Agents operate over a given subset of

information

· deliniated by ontology, tagging

and the semantic area of that

agent

· the agent interprets that

information according to its

semantic grounding

agent1

semantic domain

instantiated information

conforms to

grounding

uses



Two agents “communicate” if there is

intersection between the information

they are using

• here lies a problem

· there two agents might

interpret the information in

completely different ways

· chaos and nonsense might

result

agent1

agent2





they are using

• sensible communication only results

if the semantic domains of the agents

are aligned sufficiently

· we do not have good definitions

nor metrics to define “sufficient

enough”

· standardisation....

agent1

agent2





they are using

• harmonious communication and

understand is only achieved when the

semantic domains are identical

·hard to guarantee

·standardisation again....

agent1

agent2


Theoretical Underpinnings – Ontology Evolution

Ontologies may be given but...

· Individual tagging

· Folksonomies

· Standardisation

· Informal and implicit

· Formal and explicit

· Ontology

· Ontology emergence

· Semantic Grounding

·semantic evolution, change and

emergence

increasing formality

loca

l g

lob

al

personal tags

folksonomy

ad hoc ontology

standardised ontology

semantic strength


Theoretical Underpinnings - Logic

Description logics, normality,

soundness, completeness, decidability

and monotonicity are not sufficient

•Information needs to be removed

•Not all agents “think” in the same

way

•Logic varies according to ontology

and semantics

•Unknown values not always

interpretable as undefined

Logics will vary according to space

and even be modified on a per-agent

basis

•areas of research:

·non-monotonicity and

defeasilibity

·multi-valued logics

·consideration and interpretation

values such as

·non-insistence of completeness

and decidability

·etc


Theoretical Underpinnings – Belief, Truth, Consistency

Assertion of information by an agent

does not imply truth.

agent1 asserts

Finland Stockholm

capital


Theoretical Underpinnings - Belief, Truth, Consistency

We do not enforce consistency

according to ontology

· some spaces might...

Another agent might add additional,

contradictory information

· this might be its intent

· interpretation is left to the

reader

agent1 asserts

Finland Stockholm

capital

which conforms to:

City Country capital 1



Assertion of information by an agent

does not imply truth.

Agents 2 and 3 can interpret this

according to their beliefs and make

decisions accordingly...

...however mixed they are...

agent1 asserts

Finland Stockholm

capital

agent2

agent3 capital(Finland)=

“Stockholm” capital(Finland)=

“Stockholm”

is wrong!

capital(Finland)=

“Helsinki”



We do not enforce consistency

according to ontology

· some spaces might...

agent1 asserts:

Finland Stockholm

capital

agent2 asserts:

Finland Helsinki

capital

Finland Stockholm

capital

Helsinki

capital

giving:

which does not conform to:

City Country capital 1



While no decision procedure exists to

conclusively choose an answer, there

are options.

·Agents may employ belief revision

and truth maintenance algorithms to

clean-up such information not

adhering to known ontologies

·but this is not always desirable.

Finland Stockholm

capital

Helsinki

capital

agent3 interprets locally,

with the possible answers:

1. Stockholm is the capital of Finland

2. Helsinki is the capital of Finland

3. Both Stockholm and Helsinki are the capitals of Finland

4. Error

5. Unknown

6. Undefined

7. X is the capital of Finland, where x is not Stockholm nor Helsinki but some answer or interpretation that Agent3 wants to give


Theoretical Underpinnings - Spaces

Individual graphs of information are

localised as spaces

this is the partitioning of the “Giant

Global Graph” concept into more

localised and personal spaces.

type

type type

tag tag

tag

tag

type

type type

tag tag

tag

tag

type

type type

tag tag

tag

tag


Theoretical Underpinnings - Spaces

Individual graphs of information are

localised as spaces

each space may contain its own set of

reasoning capabilities and logic for

processing the given information

type

type type

tag tag

tag

tag

type

type type

tag tag

tag

tag

type

type type

tag tag

tag

tag


Theoretical Underpinnings – Agents and Spaces

An agent may connect simultaneously

to many spaces in order to gather the

information it needs to reason over

type

type type

tag tag

tag

tag

type

type type

tag tag

tag

tag

type

type type

tag tag

tag

tag


Theoretical Underpinnings – Agent-Space Membership

Demarcation of Spaces according to

local policy to restrict agent access

Demarcation can be potentially a

combination of:

• agent identity

• user identity

• location

• temporal characteristics

• keys (traditional security)

• etc...

type

type type

tag tag

tag

tag

type

type type

tag tag

tag

tag

type

type type

tag tag

tag

tag


Theoretical Underpinnings – Agent Mobility

Agents are mobile by way of links to

spaces

·cf: pi-calculus notions of mobility

·agents are mobile amongst

spaces

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g



Agents are atomic entities which

execute on a single device at a time

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g



Agents are atomic entities which

execute on a single device at a time

agent exist through spaces

· current implementation does

not admit mobility of executable

code, but...an agent may save its

state to a space which another

agent might use

· agent existence persistence

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g


Theoretical Underpinnings – Space Structure

Spaces “just exist”

• represented by one or more

Semantic Information Brokers (SIB)

• which are “totally routable”

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g


Theoretical Underpinnings – Space Structure

Spaces “just exist”

• each space contains (cf: architecture)

·connectivity functionality

·information storage

·full, partial or even none!

·query distribution and information

store synchronisation

·deductive closure calculation

mechanisms

•agent always gets a single, consistent

view of all information

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g

t

y

p

e

t

a

g

t

a

g

t

y

p

e

t

a

g


Theoretical Underpinnings – Spaces and Devices

Nominally a SIB executes atomically on

a single device

A device may host any number of SIBs

· even ones representing the same

space

·SIBs may have different storage

and processing capabilities

depending upon the hosting

device

·the capabilities of a space is given

by the union of all the capabilities

of the individual SIBs representing

that space

t

y

p

e

t

y

p

e

t

y

p

e

t

a

g t

a

g

t

a

g

t

a

g


Theoretical Underpinnings – Device Abstractions

Because applications emerge from

agents and spaces emerge from

SIBs we abstract the traditional or

legacy notion of application

completely from its physical

presence in any device

· even within the UI the

composition of an application

is abstracted away from the

agents themselves

} emerges from


Implementation and Distribution

•Example Demo Setup •Python

·Python 2.5.1 under Linux, Unix, Windows, Symbian etc

•C

·Linux (N800/N810)

•OpenC

·Symbian (Nokia N and E series devices)

•Java (4Q08, 1Q09)

·J2ME, J2EE

Simple XML based protocol – specification and reference implementations will be released as open source distribution 4Q08/1Q09 (estimated)


References

• Oliver, Honkola (2008) Sedvice: A Triple Space Computing Exploration Environment.

Tripcom Workshop, Galway, April 2008

• Oliver, Honkola (2008) Personal Semantic Web Through A Space Based Computing

Environment, MSW @ ICSC08, Santa Clara, August 2008 (arxiv.org: 0808.1455)

• Oliver, Honkola, Ziegler (2008) Dynamic, Localised Space Based Semantic Webs,

WWW/Internet Conference, Freiburg, October 2008

Forthcoming:

• Space Based Semantic Webs, Journal of Semantic Computation, Sept’08

• Semantic Computation, Journal of Semantic Computation, Dec’08


Current Research

•Security

•Policy

•Trust

•Ontology Construction

·tagging, folksonomies

·ontology evolution

·information recycling

·semantics

•Synchronisation and Co-ordination of agents

•Connectivity Solutions

·legacy integration

•Reasoning

·non-monotonic logics

·description logics

·planning, AI ...

•Application/Agent Construction

·tool environments

·verification/validation strategies

•Distribution

·query distribution and optimisation

·distributed deductive closure calculation


The End

towards the dynamic semantic web

Technology